# The Deep Pi, held

> **THE READING PLAQUE.** *The Deep Pi, held*, draft 0.2 — **made readable on the SX
> surface by its covering persona's reading grant; a draft, not the published form.** The
> grant stands by the operator's word (2026-09-16), verbatim: "ok - now do the same for the
> dpx." Covering: **starl3n** · endorsing: **Link Digital** (the standing pair). The
> Foundations · 8: the eighth reading of the foundations series and the first of its
> second tier — an experiment of the observatory node, held on this node. The
> construction record stands in the open workshop at
> [/sx/workshop/the-dpx-held](/sx/workshop/the-dpx-held); the experiment's own door,
> [opendata.ai/dpx](https://opendata.ai/dpx), wears the word *grown · held* and links back
> here while this grant stands. Reading is free; the way across is a statement of intent
> ([/sx/matters](/sx/matters)). This file's sha256 is pinned in the reading register.

*Draft state — this block records the state of the draft; the reading begins after the
rule.* Draft 0.1 — 16 September 2026. Author of record: Steven De Costa; composed at his
word from the experiment's record. Standing: a reading, stated never asserted; the
record it reads stands whole without it. The later locks and the walk land in a further
draft. Draft 0.2 (16 September 2026) places 3 figures and 1 table at the operator's word, each drawn from the record's own numbers and captioned with its source; the screenshots the operator will supply are not yet placed.

---

## What the Deep Pi is

Take the decimals of π after the point and cut them into blocks of 192. Place each block
on a ring of 192 vertices — a lap — with the block's last digit on vertex 0, so that the
walk of the digits leaves vertex 0 and lands back on it as the lap closes; that closing
digit is the ring's pinch. Then apply the Ducci map, which replaces every vertex's value
with the absolute difference from its neighbour, and watch what each lap does. That is
the whole instrument. Words of 256 digits on a ring of 256 are read the same way.

Two elementary results frame every measurement, and the note that proposed the
experiment claims no novelty for either. A run of k equal digits becomes k − 1 zeros
under one step and erodes one cell per step from its right end, gone after k − 1 more.
And on the 192-ring every eventual period divides 192, which is 2⁶ · 3; equality at 192
is a genericity, not a theorem. The first eight laps were measured by hand before any
rung ran: every one locked at period exactly 192, seven of them at step 64, and one at
63 — lap 4, whose last six vertices carry the six 9s at decimals 762 to 767, the Feynman
point, ending on decimal 768, which is 4 · 192 and 3 · 256 at once, the first common
boundary of the two rings. A hundred and twenty random digit seeds all locked at 192 as
well; four word seeds drained to zero in 255 or 256 steps, as a power-of-two ring
guarantees they eventually must.

From that note came the proposal: a rung ladder of 64, 128, 256, 512 and 1024 laps, with
five hypotheses registered before each rung — every period divides 192 as the control,
every lap locks at exactly 192, every lock falls at step 63 or 64, the counts of laps
with long born runs fall inside exact binomial intervals fixed at registration, and early
locks are not associated with long born runs beyond chance. A registration record is
written before the run and a results record after it, each pinned by its hash, and the
archive, not the experiment host, is the citable object. The falsification standing was
stated first: a single lap breaking the period or the lock window would be the headline,
and continued confirmation the boring outcome, to be reported as exactly that.

![The instrument: a block of 192 decimals laid on a ring of 192 vertices with its last digit on vertex 0, the pinch, and one Ducci step on a run of equal digits — k equal digits becoming k − 1 zeros, then eroding one cell per step from the right. A schematic; the ring is drawn with fewer vertices than it has.](/sx/assets/foundations/the-dpx-held/dpx-ring.svg)

## The build and the first rungs

The runner computes π twice, by Chudnovsky binary splitting and by the Gauss–Legendre
mean, and refuses to proceed unless the two agree digit for digit. The order of acts is
enforced fail-closed: a rung cannot be registered twice, cannot run without its
registration, and cannot run twice; each registration embeds the frozen model with its
seed, the exact intervals for the fourth hypothesis, the code's own hash, and the
declaration that nothing has been peeked. The first run crashed at the platform's cap on
converting large integers to strings, before any digit or fate was read, so the rung-64
registration stood valid and unrun and the runner was fixed — and the lesson entered the
record: test past platform caps at production scale.

Then the operator's runs at 128 and 256 laps, both registered before either ran, gave the
headline the proposal had said would be reported first. The lock window was refuted
twice. Locks spread from 56 to 64 — 45 of 256 laps below 63, the deepest at step 56 on
lap 43 — so the eight-lap window had been too narrow. The period held, 256 for 256, every
lap locking at exactly 192. The model rates held. And the association test answered the
note's first open question: early locks and long born runs are not associated; lap 4's
early lock was unremarkable, a rhyme retired by its own instrument. The experiment then
took the open form of the house — the door /dpx with the refutation on its face, the
ladder of registrations and results per rung, and the acts as buttons in lawful order.

![Where the 256 laps lock: steps 56 to 64, 149 laps at 64 and 62 at 63, 45 of 256 below 63, the deepest at step 56 on lap 43 — the eight-lap window refuted, every lap at period exactly 192. Measured: the rung's results as pinned in the state.](/sx/assets/foundations/the-dpx-held/dpx-locks.svg)

## The approach and the arrivals

The extension rungs left the Ducci ladder to read the decimals themselves, each rung
registered before its read with knives that land PASS, REFUTED or VOID exactly as
defined, a refutation reported first, and the decimals computed by two independent
methods that must agree on the whole territory or the rung is void.

The approach rung read laps 4,097 to 16,384 on the operator's word — "let's run this and
predict there 'is' a geometry to pi that is 100% deterministic" — with the geometry
carried as his reading and never as a knife. What was already seen was declared: eleven
runs of six or more through lap 4,096, the seven 3s at decimal 710,100 on lap 3,699,
which is 137 · 27. The toll on 137 was refuted: the last long run in the territory, six
9s at decimal 3,062,881, falls on lap 15,953, which is not a multiple of 137. The
approach ran free, as staked: no run in the last 256 laps. And the seven rhymes with
three was refuted: the first seven in the territory is seven 9s, on a lap divisible by
three but not the digit three. A declared look at lap 7,398, fifty-four times 137, found
nothing longer than four, and was a look, not a knife.

The arrivals rung took the operator's prediction of the folds — "six 9's, then the next
epoch fold of seven 3's, then eight 2's and a final string of ten 6's" — with his own
uncertainty about the order kept beside it, and his rule for where the next fold would
sit before its pinch. The first eight is eight 4s, at decimal 22,931,745 on lap 119,437,
so the eight 2s were refuted; its distance to the pinch was 159 against the rule's 139 or
181, so the position rule was refuted; and the first ten lay beyond the fifty million
decimals two methods could lock that night, so the ten was void. A single-method look to
a hundred million was disclosed as a look: ten runs of eight or more, the longest nine,
no ten. The record also marks that the rule's fit to the seven 3s was a fit, not a
prediction, since that position was known before the rule was stated.

## The band

The band rung read four hundred million decimals by two methods. The operator's words
opened it — "I deeply suspect there are no significant 'pinch points' that drive a run
of an integer greater or equal to 11 as a series" — with his likeness of π's walk to
memory that cannot turn, against the primes' knight's walk, recorded as a likeness. The
house cut three knives from his reasoning and he adopted them, naming the digit himself:
"H7 will be, I think, a series of 6's."

The first exact ten arrives inside the territory, and no eleven before it: ten 6s at
decimal 386,980,412, on lap 2,015,523, beginning four digits before the lap's pinch. So
the band with teeth passed, and ten sixes passed — the digit he had named before any
read past a hundred million. The third knife, that no long run ever crosses a pinch,
was refuted by that same run: of thirty-seven runs of eight or more in four hundred
million decimals, the ten 6s is the only one that contains a pinch, straddling it four
to six. The joint null for all three had been stated in advance at about six in a
thousand; two passed, and the one that failed failed on the very run that carried the
other two.

![The first exact ten in four hundred million decimals: ten 6s at decimal 386,980,412 on lap 2,015,523, beginning four digits before the lap's pinch and straddling it four to six — the only run of eight or more, in thirty-seven, that contains a pinch. The neighbouring digits are not drawn.](/sx/assets/foundations/the-dpx-held/dpx-ten-sixes.svg)

## The constants and the bases

Does the pinch structure belong to π, or to the ring and the fold? The next rung read e,
√2 and φ to four hundred million digits through the same ring, each with an exact
identity as its second lock, and π in bases 2, 6, 12 and 16 at matched information. No
constant produced a ten at four-to-six; only e produced a nine that straddles a pinch,
nine 1s at decimal 110,702,779, five digits before its pinch, recorded as a fact and
staked as nothing. The fold words' early drain held for φ and failed for √2 and e. In the
bases, the rarity-matched fold straddled the pinch nowhere, and the ring showed no
preference for the bases whose factors match it. The ten 6s on the pinch is π's own, at
that one place, and the record found it nowhere else it looked.

## The fates, the wave and the word

Two rungs asked whether the fold laps — the laps that open a first-of-length run — are
special in their dynamics rather than their digits. The two fold laps already run had
locked early, 63 and 62, against a mode of 64; that was the motivating observation,
marked post hoc. The three unread fold laps were run and locked at 64, 63 and 64: fold
laps lock like any lap, and both knives were refuted against a null computed over 2,999
ordinary laps beforehand. Then the eternal wave: not all five fold laps ride the outer
waves of eternal weight, but three of them do, which passed the weaker knife and failed
the stronger. And the 256-word that holds each fold drains early — 255, 255, 255, 255 and
254 against a mode of 256 — but not clearly early, so the first drain knife passed and
the second was refuted. A fold lap is a place in the digits; the ring does not feel it.

The extension rungs, their knives and their verdicts, refutations first:

| rung | territory | the knives, and how they landed |
|---|---|---|
| the approach | laps 4,097 to 16,384 | the toll on 137 refuted; the approach ran free, passed; the seven rhymes with three refuted |
| the arrivals | to fifty million decimals | the eight 2s refuted (eight 4s at decimal 22,931,745); the position rule refuted (159 against 139 or 181); the ten void |
| the band | four hundred million decimals | the band with teeth passed; ten sixes passed, the digit named beforehand; no long run crosses a pinch refuted by that same run |
| the constants and the bases | e, √2, φ to four hundred million; π in bases 2, 6, 12, 16 | no constant produced a ten at four-to-six; the fold words' early drain held for φ, failed for √2 and e; no base preference |
| the fold fates | the three unread fold laps | fold laps lock like any lap — 64, 63, 64 — both knives refuted against a null over 2,999 laps |
| the wave and the word | the five fold laps and their 256-words | three of five ride the outer waves — the weaker knife passed, the stronger failed; the words drain early but not clearly — one knife passed, one refuted |

## The firewall

No claim about π's normality or the distribution of its digits, and none over π's own
laws; its theorems stay the discipline's. No claim over the prime numbers' own laws,
ever; the operator's reading of the primes is his and is tested by nothing here. No
claim on any Millennium Prize problem, ever. No claim about physical cosmology, quantum
mechanics, consciousness or phenomenal experience. The memory likenesses are likenesses.
The operator's runs count nothing in the care census: acts, never eyes. And the geometry
of π he predicted stands or falls by his judgment of the landed record, never by a
verdict — the knives establish only whether exact statements about lap indices hold on
decimals no one had read.

## Where it sits

An applied campaign of the Number Theory set on the observatory node: the door is
[opendata.ai/dpx](https://opendata.ai/dpx), state *closed* — the record replays
read-only, the ladder's remaining rungs stand as the operator's own acts on the page,
and nothing a reader does there enters the record. Its companions on this node are the
Feynman Nines, the reading that began with the six 9s, and *The Quarter, Three Ways*,
which carries these refutations first. The registrations and results are pinned by their
hashes in the repository; the archive was designed as the citable object, the archive
act is the operator's own, and no external anchor is recorded for the Deep Pi — this
reading says so rather than borrowing one.

## Standing

Held on this node as fact until someone stands up in the record to refute it; a
refutation enters as a statement of intent at [the matters of note](/sx/matters) and
goes to the record's own path. What has since been built on the Deep Pi, and the walk a
reader may take, are the operator's hand and land in a further draft.